For operators running AI training clusters, private 5G cores, or high-performance computing racks, the awkward question is rarely about features — it is whether the provisioning layer will still be maintained when the servers beneath it are halfway through their depreciation cycle. Canonical says its Metal-as-a-Service (MAAS) bare-metal automation platform will receive up to 15 years of security maintenance, a figure the company framed around capital investments planned over multiple hardware generations.

The claim, stated in a post on Ubuntu's official blog, arrives alongside MAAS 3.8, which Canonical bills as the project's first LTS release. The announcement also details new functionality in that release, including OIDC/SSO integration, relationship-based access control (ReBAC), and deployment-time scripts. The headline number targets the mismatch between hardware and software lifecycles: hardware in AI supercomputing, telecom, and HPC deployments is typically justified over a decade or more, while the software used to image, mount on racks, and re-provision those machines is often replaced or re-architected far sooner. Each swap forces a window in which the node-onboarding pipeline is in flux — a real operational cost for stretched infrastructure teams.

What the number does and does not say

The up-to-15-year figure is Canonical's own commitment, and several important details are not addressed in the announcement. No per-release support matrix is included, so it is unclear whether the term applies per Ubuntu LTS release, as a project-level horizon, or both. It is also not stated whether the snap-packaged MAAS server component is covered alongside Ubuntu LTS itself, or on a separate lifecycle. Lifecycle tiers — full maintenance rather than security-only patches — have not been spelled out, and those distinctions determine whether the number has operational meaning or is simply a branding figure.

Separately, MAAS has been in development since around 2012. That project longevity is not the same as per-contract commercial support, which binds only organizations that hold an Ubuntu support agreement. Buyers without such a contract get no contractual horizon at all.

Competitive context

In the bare-metal provisioning field, the alternatives are largely upstream projects: Foreman, OpenStack Ironic, Metal3, and Cobbler each offer capable tooling but carry no contractual support commitment. VMware's ESXi, whose lifecycle picture has been a moving target since the Broadcom acquisition, makes lifecycle predictability a procurement question in itself. Against that backdrop, a published support term is the one differentiator that does not show up on a feature checklist — and Canonical's terms only bind customers who sign up.

That distinction is a general procurement consideration for anyone evaluating long-lived infrastructure programmes. No regulator or enterprise has publicly tied its planning to this announcement, and the development is not regional in scope.

The buyer question

The useful test when evaluating any bare-metal automation vendor is not what the tool demonstrates today, but what it commits to in year twelve. The questions worth putting to the company are which per-release matrix applies, how lifecycle tiers are split, whether the management server component is covered, and which contract clause delivers the promise. A support horizon is worth precisely what the agreement behind it says.

Canonical's up-to-15-year maintenance commitment for bare metal automation with MAAS is a claim worth investigating on those terms. Until the fine print is published, it remains a vendor statement, not a verified timeline.


對於運營 AI training cluster、私有 5G core 或高性能運算機架的營運商來說,尷尬的問題很少涉及功能——而是當底下的伺服器已用到折舊週期的一半時,provisioning 層是否仍會獲得維護。Canonical 宣布其 Metal-as-a-Service(MAAS)bare-metal automation 平台將獲得最長達 15 年的安全維護,該公司以橫跨多個硬件世代的資本投資計劃來框架這個數字。

這項聲明在 Ubuntu 官方 blog 的一篇帖文中提出,並與 MAAS 3.8 同步公布——Canonical 將其定位為該項目的首個 LTS 版本。公告同時詳述該版本的新功能,包括 OIDC/SSO 整合、基於關係的存取控制(ReBAC),以及 deployment-time scripts。這個核心數字針對的是硬件與軟件生命週期之間的不對稱:AI 超級運算、電訊及 HPC 部署中的硬件通常以十年或更長時間來攤銷,而用於為這些機器寫入 image、裝上機架及重新 provisioning 的軟件,往往遠早便被更換或重新架構。每次更換都會造成一個 node-onboarding pipeline 處於動盪的窗口期——這對人手緊張的基建團隊而言是實際的營運成本。

這個數字說明了什麼,又沒有說明什麼

最長 15 年這個數字是 Canonical 自己的承諾,公告中沒有處理若干重要細節。公告未有附上按版本劃分的支持矩陣,因此尚不清楚該期限是按每個 Ubuntu LTS 版本計算、作為項目層面的時限,還是兩者皆是。snap 打包的 MAAS server 元件是否連同 Ubuntu LTS 本身一併涵蓋,抑或按獨立的 lifecycle 處理,公告亦未有說明。Lifecycle 層級——全面維護而非僅安全修補——未有列明,而這些區分決定了該數字是否具有營運意義,還是僅為一個品牌宣傳數字。

另一方面,MAAS 自約 2012 年起一直在開發中。這種項目層面的持久性並不等同於按合約提供的商業支持,後者只約束持有 Ubuntu 支持協議的機構。沒有此類合約的買家完全得不到任何合約保障的時限。

競爭環境

在 bare-metal provisioning 領域,替代方案主要是 upstream 項目:Foreman、OpenStack Ironic、Metal3 及 Cobbler 各自提供功能完備的工具,但均不帶合約支持承諾。VMware 的 ESXi 自 Broadcom 收購以來,其 lifecycle 情況一直不斷變化,使 lifecycle 可預測性本身成為採購時的考慮因素。在此背景下,公布的 support 期限是唯一不出現在功能清單上的差異化因素——而 Canonical 的條款只約束簽約客戶。

這項區別對任何評估長期基建項目的人而言,都是一項一般性的採購考慮因素。目前尚無監管機構或企業公開將其規劃與此公告掛鈎,該發展亦不具地區性範圍。

買家的問題

評估任何 bare-metal automation 供應商時,有用的測試不是該工具今天展示什麼,而是它在第十二年承諾什麼。值得向公司提出的問題包括:適用哪一個按版本劃分的矩陣、lifecycle 層級如何劃分、管理伺服器元件是否受保,以及哪一項合約條款落實該承諾。一個 support 時限的價值,精確地等於其背後協議所載的內容。

Canonical 為 MAAS bare-metal automation 提供最長 15 年維護承諾的說法,值得按這些條件去查證。在細則公布之前,它仍然只是一份供應商聲明,而非已驗證的時間表。

新聞來源 / Original News Source